Room: AAPM ePoster Library
Purpose: build and validate a multi-omics prediction nomogram of lymph node (LN) metastasis in esophageal squamous cell carcinoma (ESCC).
Methods: ESCC patients who underwent extensive LN dissection in our hospital were included. The basic information and postoperative pathology results were collected. 239 radiomics parameters of CT arterial phase images were acquired prior to preoperative use of LIFEx 4.00. qRT-PCR was used to detect the expression of miRNA-21-5p, miRNA-143-5p, and miRNA-145-5p in paraffin-embedded tumor tissues. The patients were divided into a training group(n=150) and a test group(n=64). In the training group, LASSO regression was used to screen radiomics features and to establish radiomics signature. The univariate analysis was used to screen for clinicopathological factors and gene expression. Radiomics signature, significant clinical pathological factors and genes were included in the multivariate logistic regression; and the nomogram was drawn. Calibration curves, decision curves and receiver operating characteristic curves (ROC) analyzed the availability of the nomogram.
Results: regression screened nine radiomics features that were significantly associated with LN status. The nomogram contains the radiomics signature, the LN status determined by CT, and miRNA-145-5p expression status. The ROC shows that the area under the curve of nomogram predicting LN metastasis status was 0.905 for the training group and 0.896 for the test group. The decision curve shows that the model is clinically usable.
Conclusion: established multi-omics nomogram including CT LN status, radiomics features, and miRNA-145-5p expression may facilitate the prediction of LN status of ESCC before treatment.